LiftLog plans can be exported to, and imported from, .liftlogplan files. A plan file is plain JSON, so you can write one yourself, share one with a friend, keep one in version control - or have an AI write one for you.
To import a plan, either tap a .liftlogplan file on your device (LiftLog is registered to open them), or open the app and go to Plans -> Import.
To export one, open Plans, tap the ⋮ next to the plan and choose Export to file. LiftLog writes the .liftlogplan and hands it to the system share sheet, so you can save it to Files or Drive, AirDrop it, or mail it to yourself - whatever gets it somewhere you can reach it again.
The authoritative definition of the format is the JSON Schema at docs/schemas/program-blueprint/ProgramBlueprint.json. It is generated from the app's own models, so it is always in step with what the app will accept.
This repository ships a Claude skill that writes plan files for you. Describe the training you want - "a 4-day upper/lower split for an intermediate lifter, dumbbells only" - and it produces a .liftlogplan file, validated against the schema above before you ever see it.
/plugin marketplace add LiamMorrow/LiftLog
/plugin install liftlog-plan-builder@liftlog
Download create-liftlog-plan.zip, then in Claude go to Customize -> Skills -> + -> Upload a skill and select it.
That zip is rebuilt whenever the skill is released, so it always matches the newest released format rather than whatever is on main.
Once it is installed, just ask for what you want:
Build me a 5x5 strength program, three days a week.
Get the resulting file onto your phone (AirDrop, email, or save it to Files) and tap it.
The skill reads plans as happily as it writes them, so you can send it the one you are training on rather than describing it. Export the plan from Plans -> ⋮ -> Export to file, save it somewhere you can get at it from your computer (Drive, or mail it to yourself), then give the .liftlogplan to Claude and say what you want changed:
Here's my current plan. Swap all the barbell work for dumbbells, and add a fourth day.
This plan is taking too long. Cut it to 45 minutes without losing the compounds.
Have a look at my plan and tell me what's missing.
Exporting means Claude sees your real exercise names, rest times, and progressive overload settings, so what comes back is your plan with the change made - not an approximation of it rebuilt from a description.
Importing the result adds a new plan; it does not overwrite the original. Delete the old one from Plans once the replacement is in. Your logged workout history is kept separately and is untouched by either.
Nothing about the format is Claude-specific. To use ChatGPT, Gemini, or anything else, give it the schema and the rules below:
Write me a LiftLog workout plan as a single JSON object matching the schema at https://github.com/LiamMorrow/LiftLog/blob/main/docs/schemas/program-blueprint/ProgramBlueprint.json
Rules that are easy to get wrong:
- Treat every field in the schema as required, apart from the handful marked optional. Empty strings for
notesandlink."version": 3on the root object,"version": 6on every session.- Weights, distances and progression steps are decimal strings:
"2.5", not2.5. Rep counts are plain integers.- Rests and cardio times are ISO-8601 durations:
"PT3M","PT90S".- A weighted exercise has no set count: one entry in
plannedSetsis one set. Cardio usessets, an array of set objects.- Supersets are a flag on the preceding exercise:
"supersetWithNext": true.typevalues are case-sensitive. Exercise types are PascalCase ("WeightedExerciseBlueprint"); cardio targets, progression scopes and resistance are lowercase or camelCase ("time","allSets","bodyweight").[describe the training you want here]
Save the result as My Plan.liftlogplan.
A plan file is one JSON object: a name, a date, and a list of sessions. Each session is a training day holding a list of exercises.
{
"version": 3,
"name": "Push Pull Legs",
"lastEdited": "2026-07-12",
"sessions": [
{
"version": 6,
"name": "Push",
"notes": "Chest, shoulders and triceps.",
"exercises": [
{
"type": "WeightedExerciseBlueprint",
"name": "Barbell Bench Press",
"plannedSets": [
{ "reps": { "min": 5, "max": 5 } },
{ "reps": { "min": 5, "max": 5 } },
{ "reps": { "min": 5, "max": 5 } }
],
"restBetweenSets": { "minRest": "PT3M", "maxRest": "PT5M", "failureRest": "PT5M" },
"supersetWithNext": false,
"notes": "",
"link": "",
"resistance": "external",
"progression": [
{
"axis": "load",
"step": "2.5",
"scope": { "type": "allSets" },
"trigger": "allSetsMetTarget"
}
]
}
]
}
]
}Complete examples live in plugins/liftlog-plan-builder/skills/create-liftlog-plan/examples/ - one weighted plan and one cardio plan.
An exercise is either a WeightedExerciseBlueprint or a CardioExerciseBlueprint, chosen by its type. The two can be mixed within a session.
Weighted exercises have a list of planned sets, rest times, a resistance, and a list of progression rules. plannedSets holds one entry per set, each with that set's rep target as a min/max band - min === max is a plain "five reps". restBetweenSets needs all three of minRest, maxRest, and failureRest (the last being the rest taken after missing a rep target). link is a URL explaining the movement, and should stay "" unless you have a real one.
resistance says where the load comes from: external for barbells, dumbbells and machines (the logged weight is the weight lifted), bodyweight for pull ups and dips (the logged weight is what is added on top of the lifter), or none for movements like crunches, where there is no weight at all and reps are the whole story.
supersetWithNext is how supersets are expressed: there is no superset group. Setting it to true means "perform this back-to-back with the next exercise in the list, without resting".
Cardio exercises have no resistance and no progression. Their sets is an array - one entry per interval - and each entry has a target (either a time or a distance) plus six track* booleans controlling which fields the app shows you while logging, and optionally its own restBetweenSets.
progression is an ordered list of rules. After a session where every set met its target, the first rule that still has room to move is applied - and only that one. An empty list means the exercise never moves on its own.
| Field | Effect |
|---|---|
axis |
What the rule moves: load or reps. |
step |
How much to add, as a decimal string. Unitless - "2.5" is the same pair of small plates in a kilo gym and a pound gym. |
scope |
{ "type": "allSets" }, or { "type": "lowestSets", "pick": … } with first, middle, last or all to raise only the lowest. |
trigger |
Always allSetsMetTarget. |
ceiling |
Optional, reps only. Where the rule stops, handing over to the next one. |
onCeiling |
Optional, reps only. reset drops the reps back to the plan's when a later rule takes over. |
Only a reps rule takes a ceiling, so a load rule never runs out of room and must come last - anything after it can never fire. Reps to a ceiling followed by weight is double progression: the reps climb, then the weight goes up and the reps start again.
See Progression for how the rules behave in the app.
The skill ships a validator, and you can run it directly on any file. It needs no dependencies and no network:
node plugins/liftlog-plan-builder/skills/create-liftlog-plan/scripts/validate-plan.mjs "My Plan.liftlogplan"It lists every problem at once, with the path to the offending field:
My Plan.liftlogplan is not a valid LiftLog plan:
plan/sessions/0/exercises/0/restBetweenSets/minRest must match format "duration"
plan/sessions/1/exercises/0/progression/0/step must be string
This is worth doing, because when a field is wrong the app will only tell you "That file isn't a valid workout plan" - it cannot tell you which one.
Both the schema and the validator are generated from app/src/models/storage/versions/latest/blueprint.ts. After changing that model:
cd app && npm run json-schemaThis rewrites the published schema, the copy bundled into the app, the copy inside the skill, and the skill's validator. CI checks the committed copies still match the models on every branch, so all four move together on main. Users only see the change once the plan-builder-skill tag advances, which scripts/release-plan-builder-skill.ts does - run by hand once the release is live in the stores, not when it is uploaded to them.
See Storage Migrations for the full process of changing a stored model.